Anthropic announced Friday that it has chosen Accenture and the consulting firm’s AI unit Faculty as its first embedded evaluator. The arrangement supports the initial element of a three-step plan outlined by Anthropic CEO Dario Amodei on September 12 to moderate the pace of advanced AI development while safety measures advance.

Amodei’s Proposal and the Case for Independent Access

Amodei’s plan centers on concerns that artificial intelligence systems are improving at an accelerating rate through recursive self-improvement, in which models help create successive generations of more capable systems. He argued that this dynamic, if left unaddressed, could advance faster than human capacity to understand and control the technology, raising the possibility of severe harm.

The first step in the proposal calls for independent evaluators granted access comparable to that of internal employees. Anthropic stated it had already committed unilaterally to this approach. The partnership with Accenture represents the company’s first concrete step to put that commitment into practice.

Under the arrangement, Accenture Faculty will evaluate and red-team models, conduct alignment assessments, and test model safeguards. Anthropic noted that embedded evaluation remains an emerging practice, and the precise operational details are still being finalized.

Investment Commitments and Funding Approach

Anthropic and Accenture each expect to invest at least $1 billion in the initiative over the next five years. Because no established mechanism currently exists to finance independent evaluation work, Anthropic said it will fund Accenture’s efforts directly in the near term, citing the urgency of the task. Over the longer term, the company indicated that pooled industry resources or government sources could provide sustained support.

The partnership is non-exclusive. Anthropic said it plans to name additional evaluators in the coming weeks.

Industry Responses to the Slowdown Call

Amodei’s September 12 proposal drew mixed reactions from technology leaders. OpenAI CEO Sam Altman and SpaceX CEO Elon Musk responded positively. Nvidia CEO Jensen Huang took a different view, arguing that regulatory measures aimed at slowing development were unnecessary.

Accenture’s Faculty unit brings experience testing and evaluating models for major AI laboratories and constructing complex systems designed with safety and ethical considerations from the outset. The collaboration positions independent evaluation closer to the model development process rather than treating it as a purely external review.

As the program takes shape, attention will focus on how employee-like access is implemented in practice and whether the addition of further evaluators helps establish broader independent oversight capacity across the sector.